An empty garden

I implemented over 40 automations, all via Slack and all with a Human in the loop (Im not out to replace anyone). However, user traction was dismal! This was a serious setback and a problem to be solved.

This is Part 1 of an N part series, where I try different approaches to get a Sales Team in a scaling B2B company to adopt AI. Not just to automate existing work, but to make them better Sales people.

Since August 2025 I’ve been keeping my ex-CEO, Hassan, abreast on the work I’ve been doing. When I told him Im pivoting Yellow Intelligence into a Sales AI transformation consultancy he offered to be my first client. He introduced me to Rafay, the company’s newly appointed and already highly impactful Chief Business Officer.

Rafay showed me the impressive work he had already been doing with Claude Cowork. He had seen firsthand what AI agents could unlock and wanted to extend that productivity across the wider team. That’s where I came in.

I first got down to replicating and scaling some of the AI workflows that he had designed for his personal use. We did this using a multi-tenant openclaw, with each user assigned to a dedicated sub-agent. The main agent behaved as a ceo. This allowed me to create centrally adminsitrated skills and workflows while giving users the flexibility and privacy to create their own.

We connected google workspaces which was a big source of intenral and external information (including call notes). Hubspot was integrated as a current system of record. Slack to consume internal conversations and finally granola for call notes that were missed by Gemini.

We then organised this into a clean Sales Gbrain with Companies, Deals and Contacts as entities, all backlinked and embedded, so that disparate sources of information on a deal was neatly organised into a knowledge graph. We enhanced the Gbrain implementation by adding a Product Gbrain. This had the entire product catalogue, including value proposition, manuals and specs. Both these Gbrain implementations provided easy and accurate information on live deals as well as products directly via Slack. Something not otherwise available before.

On top of this we built elegant AI powered workflows that solved long standing problems. Some examples:

1. Central task maanger, that automatically created, assigned and managed tasks for the entire team automatically.
2. Automated CRM management, so sales team never had to open Hubspot dashboard to make updates.
3. Automated presentation generation, so less dependence on pre-sales teams.
4. Automatic RFP synthesis, where the AI extracted key requirements → compared it to the product catalog and provided clear actionable report and RFP response.
5. Study the complete history of a client engagement history and product catalogue to better prepare for meetings.
6. Automated followups to clients, chasing them to respond to a proposal, share a PO, clear dues or asking updates on an expected tender.

Other than this we did a ton of the things that I personally consider table stakes now; like automated emails, calendar management, slack thread synthesis etc

Despite building all of these cool faeatures, usage was was dismal. We had 7 beta users, and their median weekly session rate was 8. i.e. The median user used our beautiful and sexy slack native AI ecosystem only 8 times per week! Only 8!! Our target, was 35!!!!!

Not only did this baffle me, but it annoyed me to no end. Why were these people so ungrateful? “Not only did my system make their jobs easier, I didnt automate their jobs! Dont they know that we could have done that!”, I ranted to my cofounder. Just then, I got that light bulb moment.

Im going to automate everything the sales team does right now, or atleast as much as I can. “Hey, you know those things you did for most of the day….. well that’s been all automated now. But here’s a really powerful AI tool… see what you can do .”

The intention was never to reduce head count anyway, this team was carefully built after years of hiring and attrition. Such people can not be replaced, the relationships they have built with their clients can not be replaced.

But how can we get them to use AI to become better sales people. Handing them an AI tool and hoping they would use it did not work. Now we’re going to try dragging them by their nose.

Will this work? Lets see, its an iterative experiment, and Ill post my results in Part 2 .